Texture classification using run difference matrix

A new approach to texture classification for ultrasound liver diagnosis using run difference matrix combined with neural networks was developed. The run difference matrix comprised the gray level difference along with a distance between the pixels. Ten parameters defined from the run difference matrix were used as the input of the neural network. Because the separation between the normal and chronic hepatitis liver was difficult, two schemes of neural networks were trained independently. Scheme I is for the classification of normal, fatty and cirrhotic liver while scheme II is for the chronic hepatitis, fatty, cirrhotic and normal liver. The preliminary results showed that the run difference method was more sensitive to small structural changes than the conventional run length method and showed better separability between the diseases